Health care-associated infections (HAIs) affect one in every 20 hospitalized patients and account for $10 billion dollars in potentially preventable health care expenditures annually. Current efforts at detection of HAIs are limited to manual chart review which hinders the generalizability and scalability of HAI detection. My goal in seeking a Mentored Clinical Scientist Career Development Award is to acquire the necessary training, practical experience, and knowledge to develop a health services research career as a principal investigator focusing on leveraging novel health information technology (HIT) tools to improve the measurement of surgical health care quality, safety, and effectiveness. To continue my progress towards this goal, the objective of this project is to address the challenges of HAI detection by developing a robust and portable automated HAI surveillance toolkit. This toolkit will combine structured electronic health record (EHR) data with rich information locked in clinical notes using machine learning and natural language processing (NLP) to identify HAIs after surgical procedures. Our overall hypothesis is that combining structured variables from the EHR supplemented with NLP will improve our ability to identify HAIs after surgical procedures. To test the central hypothesis and accomplish the objectives for this application, I will pursue the following three specific aims: 1) Determine the EHR data elements indicative of postoperative HAIs and evaluate the performance of a novel HAI surveillance algorithm; 2) Identify the presence of postoperative SSIs from clinical notes using an automated portable NLP-based algorithm; 3) Apply user-centered design to create a high fidelity prototype of a surgical quality dashboard incorporating our HAI case detection methodology. This contribution is a significant first step in a continuum of research that utilizes the large amounts of data in the EHR combined with novel HIT methods to improve the measurement of surgical health-care quality, safety, and effectiveness. This approach is significant because the tools developed in this proposal have potential to serve as a prototype for identification and monitoring hospitals adverse events and could be replicated on a national scale. The proposed research is innovative in its approach using a combination of structured and unstructured data in the EHR along with novel machine learning and NLP tools to create a generalizable surveillance toolkit for the detection of HAIs. This proposal is responsive to the AHRQ Special Emphasis Notice (NOT-HS-13-011) specifically addressing the use of HIT to improve quality measurement. I have assembled a mentoring team who all internationally recognized experts with long and successful track records of funding and trainee mentorship. This project will provide the means to place me on a trajectory towards a health services research career focused on improving the measurement of surgical health-care quality, safety, and effectiveness using novel HIT tools. In summary, my previous training and experience, innovative research plan, high-quality training plan, first-rate mentorship team, and supportive research environment give me the highest likelihood of success to research independence with the proposed K08 award.
/RELEVANCE TO PUBLIC HEALTH The proposed research and career development plan are relevant to public health because health care- associated infections (HAIs) affect about one in every 20 hospitalized patients and account for $30 billion dollars in potentially preventable health care expenditures annually. The objective of this project is to develop robust and portable automated surveillance toolkit that combines structured EHR data with rich information locked in clinical notes using natural language processing to identify HAIs after surgical procedures. The proposed research has the potential to serve as a prototype for identification and monitoring of numerous hospitals adverse events and could be replicated on a national scale.
